For the first time ever, I'm writing tests before fully implementing a feature and I'm uncovering bugs in its output. This feels very weird for me, but it's definitely the way to go.